ANN-CA模型改进及城市用地动态模拟

Simulation of urban land use change based on the improved ANN-CA

针对传统ANN-CA(artificial neural network—cellular automata)模型在城市动态模拟中用地转换阈值确定繁琐以及全局模拟效果不佳的问题。以ArcGIS和Matlab2012b为开发平台,构建了一个"自上而下"和"自下而上"相结合的改进模型,以重庆市南岸区为例进行土地变化模拟,对比模型改进前后的模拟效果并讨论邻域设置对模型精度的影响。结果显示,改进后模型的模拟结果不论是精度还是Kappa系数都高于传统模型的模拟结果,而且改进后模型转换阈值的设置方便灵活,无需经过多次试验来获取,随着邻域设置范围的增大,模型的模拟精度也得到一定的改善。

In the traditional ANN-CA model, the deficiency of determining threshold of land use conversion and global simulation result in the dynamic simulation of the city land use. So, this paper takes ArcGIS and Matlab2012b as the developing platform, constructs the model of a combination ＂top-down＂ and ＂bottom-up＂, and use the Nan Jan District of Chongqing as an example for verification the model in the land use simulation. The results show that： the improved model to simulate has higher precision and Kappa coefficient than traditional; In addition, the threshold is set conveniently by the improved model; at the same time with the increase of neighborhood setting range, the simulation accuracy of the model has been improved.